Researchers at Oregon State University have created a new light-responsive electronic device that closely mimics how the human brain functions. This breakthrough could pave the way for artificial intelligence systems that operate faster and consume significantly less power.
Published in Advanced Functional Materials, this innovative device integrates three key capabilities into a single unit: it detects light, stores information, and processes signals. By combining these functions, the device has the potential to make AI operations much more efficient.
Currently, most AI hardware relies on separate components for sensing, memory storage, and computation. Data must constantly transfer between sensors, memory banks, and processing units, which consumes a lot of energy and slows down performance.
This new device approaches the problem differently by performing all three functions within one component, resembling the way the human brain processes information. Professor Larry Cheng, an electrical engineering and computer science expert from Oregon State University, explains that this technology introduces a novel way to handle data directly at the point of detection.
The device, known as a phototransistor—meaning it responds to light—generates electrical charges when illuminated and stores them internally as a form of memory. What’s particularly remarkable is that these stored memories aren’t fixed; their strength can be electronically adjusted, similar to how memories in the brain can be reinforced or fade over time.
In biological brains, chemical signals influence which memories are solidified and which are forgotten. The new device emulates this process by applying small electrical signals, allowing researchers to control the longevity of the stored information—either making it last longer or fade more quickly.
This adjustable memory duration marks a significant step in neuromorphic computing, which aims to develop computing systems that operate more like biological brains. The device is constructed by layering a light-sensitive organic material over an oxide semiconductor, which conducts electricity. When light hits the organic layer, it produces electrical charges that can become trapped inside. Even after the light is removed, these charges continue to influence the current flowing through the device, effectively allowing it to “remember” prior exposure to light.
Researchers can manipulate this memory by shifting the trapped charges closer to or farther from the pathways that electrons travel through. Bringing the charges closer amplifies their effect, extending the memory’s lifespan. Moving them away weakens the influence, causing the memory to fade faster.
The ability to program the memory’s lifetime offers a flexible window for processing visual and sensory data right at their origin, which could lead to more energy-efficient vision systems, smarter sensors, and AI technologies that think and respond more like humans.